Seediscussions,stats,andauthorprofilesforthispublicationat:https://www.researchgate.net/publication/316801394

CommunicatingPersonalGenomicInformationtoNon-experts:ANewFrontierforHuman-ComputerInteraction

ArticleinFoundationsandTrends®inHuman-ComputerInteraction·January2017

DOI:10.1561/1100000067

CITATIONS

0

READ

1

4authors,including:

Someoftheauthorsofthispublicationarealsoworkingontheserelatedprojects:

CollaborativeKnowledgeProductionViewproject

G-nomeSurfer-AtabletopinterfaceforcollaborativeexplorationofgenomicsViewproject

OritShaer

WellesleyCollege

71PUBLICATIONS1,154CITATIONS

SEEPROFILE

OdedNov

NewYorkUniversity

92PUBLICATIONS1,746CITATIONS

SEEPROFILE

LaurenWestendorf

WellesleyCollege

8PUBLICATIONS11CITATIONS

SEEPROFILE

AllcontentfollowingthispagewasuploadedbyOritShaeron15May2017.

Theuserhasrequestedenhancementofthedownloadedfile.Allin-textreferencesunderlinedinblueareaddedtotheoriginaldocumentandarelinkedtopublicationsonResearchGate,lettingyouaccessandreadthemimmediately.

Foundations and TrendsR© in Human-ComputerInteraction

Vol. 11, No. 1 (2017) 1–62c© 2017 O. Shaer, O. Nov, L. Wstendorf, and M. BallDOI: 10.1561/1100000067

Communicating Personal Genomic Informationto Non-experts: A New Frontier for

Human-Computer Interaction

Orit ShaerWellesley College, Wellesley, MA, USA

oshaer@wellesley.edu

Oded NovNew York University, USA

onov@nyu.edu

Lauren WestendorfWellesley College, Wellesley, MA, USA

lwestend@wellesley.edu

Madeleine BallOpen Humans Foundation, Boston, MA, USA

madeleine@personalgenomes.org

Contents

1 Introduction 2

2 Personal Genetics Tutorial 62.1 DNA sequencing . . . . . . . . . . . . . . . . . . . . . . . 9

3 Background 133.1 Historical context and new technologies . . . . . . . . . . 133.2 Online interaction with personal genomics . . . . . . . . . 143.3 Omic technology . . . . . . . . . . . . . . . . . . . . . . . 153.4 Genomic data privacy and access . . . . . . . . . . . . . . 173.5 Public genomes and follow-up research . . . . . . . . . . . 193.6 Characteristics of personal genomic data . . . . . . . . . . 203.7 Uncertainty in personal genomics . . . . . . . . . . . . . . 20

4 Related Work 224.1 Personal informatics . . . . . . . . . . . . . . . . . . . . . 224.2 Communicating uncertainty . . . . . . . . . . . . . . . . . 234.3 Interaction with biological data sets . . . . . . . . . . . . 254.4 DIY genome analysis . . . . . . . . . . . . . . . . . . . . 25

5 Emerging Framework for HCI for Personal Genomics 285.1 Understanding users . . . . . . . . . . . . . . . . . . . . . 29

ii

iii

5.2 Informing users . . . . . . . . . . . . . . . . . . . . . . . 345.3 Empowering users . . . . . . . . . . . . . . . . . . . . . . 41

6 Broader Implications 486.1 For HCI researchers . . . . . . . . . . . . . . . . . . . . . 486.2 For practitioners . . . . . . . . . . . . . . . . . . . . . . . 496.3 For society and policymakers . . . . . . . . . . . . . . . . 49

Acknowledgements 51

References 52

Abstract

Recent advances in genetic testing and Internet technologies have led toa dramatic increase in the access non-experts have to their own personalgenomic data. Such data are complex and sensitive, involve multipledimensions of uncertainty, and can have substantial implications onindividuals’ behavior, choices, and well-being. Personal genomic dataare also unique because unlike other personal data, which might changefrequently, genomic data are largely stable during a person’s lifetime;it is their interpretation and implications that change over time as newmedical research exposes relationships between genes and health.

Future progress in genetic research and technologies is likely tofurther increase the availability of interactive personal genomic infor-mation to non-experts. This trend raises technological, ethical, andregulatory concerns related to how people make sense of, engage with,and rely on their personal genomic data. Such concerns are not onlyof paramount importance for health professionals and policymakers,but are also a pressing issue for human–computer interaction (HCI)research. HCI tools, methods, and practices can help make genomicinformation more accessible and understandable to non-experts. Weargue that the complexity, importance, and personal relevance of thistype of information makes understanding, informing, and empoweringnon-experts’ interaction with personal genomics a key challenge thatlies ahead for the HCI community.

In this article, we explore the roles HCI can play in helping non-experts contribute, understand, engage with, and share their personalgenomic information. This article is also a call-to-action for those ofus interested in the intersection of personal informatics and HCI, and,more broadly, in facilitating non-expert interaction with large amountsof complex, personal, and uncertain information.

O. Shaer, O. Nov, L. Wstendorf, and M. Ball. Communicating Personal GenomicInformation to Non-experts: A New Frontier for Human-Computer Interaction.Foundations and TrendsR© in Human-Computer Interaction, vol. 11, no. 1,pp. 1–62, 2017.DOI: 10.1561/1100000067.

1Introduction

“A firsthand familiarity with the code of life is bound toconfront us with the emotional, moral and political baggageassociated with the idea of our essential nature. People havelong been familiar with tests for heritable diseases, and theuse of genetics to trace ancestry — the new “Roots” — isbecoming familiar as well. But we are only beginning torecognize that our genome also contains information aboutour temperaments and abilities. Affordable genotyping mayoffer new kinds of answers to the question “Who am I?” —to ruminations about our ancestry, our vulnerabilities, ourcharacter and our choices in life.” [Pinker, 2009]

Recent years have seen a dramatic growth in the availability ofpersonal genomic data to non-experts, often online and in interactiveforms [Dudley and Karczewski, 2013]. The field of personal genomicsis rapidly growing, as the cost of sequencing a human genome hasfallen from approximately $100 million in 2001 to $1,000 in 2016, arate much faster than Moore’s Law [Church, 2005] (see Figure 1.1).The Precision Medicine initiative and similar health research projectsincreasingly highlight the potential for genetic data to become a

2

3

Figure 1.1: A graph demonstrating the precipitous decline in genome sequencingcost since the completion of the Human Genome Project in 2001. Whole genomesequencing is now around $1,000 per genome, five orders of magnitude lower thanthe $100 million cost of the Human Genome Project (from the National HumanGenome Research Institute).

standard component of health records [NIH]. In the meantime, direct-to-consumer genetic testing (DTCGT) companies have made genomicinformation available to millions of individuals without the involvementof a healthcare provider through online platforms. These individuals arethereby confronted with unprecedented amount of potentially sensitiveinformation about their genetic data, which influences their decisions,emotional states, and wellbeing [Davies, 2010].

The use of Web-based interactive technologies to deliver personalgenomic information raises questions about how non-experts makesense of, engage with, and rely on their personal genomic data. Futureprogress in genetic research and technologies is likely to furtherincrease the availability of interactive personal genomic information

4 Introduction

to non-expert users. This trend raises technological, ethical, andregulatory concerns. For example, in 2013, the U.S. Food and DrugAdministration (FDA) ordered 23andMe, a direct-to-consumer genetictesting company, to stop providing risk assessment reports, statingthat “serious concerns are raised if test results are not adequatelyunderstood” [Pollack, 2013]. Providing consumer reports partiallyresumed in 2015 when 23andMe recieved FDA approval for a vettedset of gene variant interpretations [Pollack, 2015]. While such concernsare imperative for policymakers, it is also vital to consider these issuesfrom the perspective of human–computer interaction (HCI) research.

Specifically, the highly personal and dynamic nature of individually-relevant personal genomic information, which is constantly updatedbased on new research results, raises important HCI questions, includ-ing: what are the functional requirements for supporting meaningfulengagement of non-experts’ with personal genomic information? Howcan we design effective interaction with personal genomic information?How can we evaluate the effectiveness of techniques for interaction withpersonal genomic information? Can user interface design interventionsimpact users’ willingness to share their personal genomic data?

An additional important aspect of communicating personalgenomics to non-experts is the understanding of the vast common-alities and genealogical similarities among people from different racialand ethnic backgrounds. Future HCI work on personal genomics has acritical role in making these commonalities apparent to users, just asmuch as the differences.

Because the technology that enables non-experts to interact withpersonal genomic information is new, there is little HCI research onit. Given the tremendous growth in the scale and scope of personalgenomic information available to non-experts, personal genomics rep-resents a new and fast-growing frontier where HCI research can makea significant difference in people’s lives.

This article explores the roles HCI can play in helping non-expertsto understand and engage with personal genomic information. It is alsoa call to action for those of us interested in the intersection of personalinformatics and HCI, and, more broadly, the facilitation of interaction

5

of non-experts with large amounts of complex and uncertain personalinformation.

We begin with a brief tutorial of personal genetics (Section 2), fol-lowed by a description of the historical context and new technologiesof personal genomics, and a discussion of the characteristics of per-sonal genomic data (Section 3). We then survey related work fromhuman–computer interaction and personal informatics (Section 4).Section 5 presents a framework for HCI research focusing on under-standing, informing, and empowering non-expert users of personalgenomics. Finally, in Section 6 we discuss broader implications for HCIresearchers, for personal genomics practitioners, for policy makers, andfor society.

2Personal Genetics Tutorial

The term personal genetics describes the information which an individ-ual has inherited from their biological parents. An individual’s physicaltraits — features of their biology, body, and health — are partly inher-ited and often are influenced by environmental and behavioral factors.At its most fundamental level, this inheritance is discrete in nature:units of inheritance are called genes. Some traits, like rare diseases andABO blood types, are associated with a single gene. Many other traits(e.g. risk for diabetes, or height) are the products of many genes, andare also influenced by environmental factors like diet.

The biological basis of inherited traits is DNA — deoxyribonucleicacid, a linear chain of nucleic acids found in all cells. The sequence ofthese four nucleic acids, a.k.a. bases — adenine, cytosine, guanine, andthymine (A, C, G, and T) – is molecular information that drives almostall features of life. DNA is inherited, copied, and passed from parent tochild. The collective set of DNA molecules for an individual, inheritedfrom their mother and father, is their genome. Each gene is a sectionof a DNA molecule that contains information for creating a specifictype of protein, each of which affects a specific aspect of cell biology.

6

7

Figure 2.1: A “karyotype” of a human genetic male shows the genome’s 23 pairs ofchromosomes. With the exception of the sex chromosomes in genetic males (X andY), chromosomes exist as matched pairs: one copy is inherited from each parent.(Public domain, from NHGRI via Wikipedia.)

Proteins are the primary functioning units of life and play a varietyof roles: they can be structural elements, or they may drive chemicalreactions that would not otherwise occur.

Each individual human genome consists of two slightly differentcopies of 3 billion nucleic acids, organized into 23 pairs of chromosomes(Figure 2.1). As demonstrated in Figure 2.2, genes are distributed alongchromosomes and are often interleaved with nonfunctional DNA. Thefunctional fragments of gene DNA, accounting for 1% of the genome,are called exons (because this sequence is “exported” and used else-where in the cell).

Genetic variants account for inherited differences between individ-uals. However, because most of the genome is nonfunctional, mostgenetic variants have no biological effect. Those that do have an effecttypically occur in exon regions: and they have a biological effect becausethey change the protein the gene produces.

8 Personal Genetics Tutorial

Figure 2.2: Each chromosome consists of a single, densely packed DNA strand.Genes occur as segments within the DNA strand, and genetic data is encoded withinthe sequence of four possible bases at each position in the DNA chain: Adenine,Cytosine, Guanine, and Thymine (A, C, G and T). Genes are often interleavedwith nonfunctional DNA (intron regions). Genetic variants which change a gene’sbehavior typically affect exon regions — the DNA sequences used to produce agene’s protein product.

Each genome exists largely in duplicate: for most genes and DNAsegments, one copy is inherited from each biological parent. When anindividual carries a genetic variant, it could be present in both genes(homozygous, referring to the sameness of each copy) or it could bepresent in only one (heterozygous). The exception to this is the sexchromosomes: males have only one copy of each “X” and “Y” chromo-some, and their variants on these chromosomes are called hemizygous.

The biological consequences of which involve genetic variants in asingle gene can be affected by whether both copies of a gene carry

2.1. DNA sequencing 9

the same variant. Genetic variants that only have an effect when bothcopies of a gene are affected are called “recessive” (these typically havelittle to no effect when heterozygous, resulting in an individual hav-ing carrier status). While recessive variants have no immediate impacton individual health, carrier status for genetic diseases can influencereproductive decisions. In contrast, genetic variants that always havean effect are called dominant. Not all variants are clearly recessive ordominant: for example, a variant may have a weak effect when het-erozygous, and a stronger effect when homozygous. Figure 2.3 explainsthe effect of genetic variants within a single gene, also referred to asmonogenic effects.

2.1 DNA sequencing

There are various technologies for “reading” DNA molecules. Sinceinformation is contained within the molecular sequence of the DNA’snucleic acids (i.e. bases), these technologies are commonly referred toas DNA sequencing. Human genome sequences are mostly identical toeach other: sequence differences between two individuals account forjust 0.1% of all nucleic acids. These few genetic variants account formany differences between two humans.

Although sequencing the human genome cost $3.6 billion when itwas completed in 2003 [NHGRI, 2016], improvements to DNA sequenc-ing have since outpaced Moore’s Law: whole genome sequences can nowbe purchased for under $1000 (Figure 1.1) [Keshavan, 2016, Regal-ado, 2016]. Two major technologies have led to vastly increased effi-ciency in producing genetic data: high-throughput DNA sequencing, andmicroarray-based genotyping. The first technology, high-throughputDNA sequencing, is comprehensive — it enables whole genome sequenc-ing (see Figure 2.4) and exome sequencing (a lower-cost option thattargets gene exon regions for around $500 [Molteni, 2016]). The sec-ond technology, microarray-based genotyping, is less comprehensive butprovides a lower-cost option: for $100–$200 it can be used to test amillion specific locations that are known to vary between individu-als [23andMe, AncestryDNA]. Because genotyping is the lowest cost

10 Personal Genetics Tutorial

Figure 2.3: Monogenic effects involve variants in a single gene. Whether variantsin a gene have an effect on traits or health depends on how many copies of thegene have variants, and on the variant’s inheritance pattern: Dominant, Recessive,X-linked, or Additive.

option, it is currently used by leading direct-to-consumer genetic test-ing companies like AncestryDNA and 23andMe.

Personal genetic information gained through the process of genotyp-ing or high-throughput DNA sequencing can be used for two seeminglydisparate purposes: ancestry and biological traits. Ancestry information

2.1. DNA sequencing 11

Figure 2.4: Current technology for high-throughput DNA sequencing starts with abiological sample (e.g. blood or tissue). A parallelized sequencing process generatesmillions of short “reads”. Reads are then computationally assembled to determine anindividual’s whole genome sequence [from the National Human Genome ResearchInstitute].

can be discovered because DNA is inherited: similarities between indi-vidual genomes can be used to discover near relatives (e.g. cousins),as well as predict more distant ancestry (i.e. geographical origins andassociations with race/ethnicity). But because DNA is responsible forbiological function, it can also be interpreted to understand biological

12 Personal Genetics Tutorial

traits. This health-relevant aspect of genome information has been sub-ject to regulatory concerns (e.g. by the United States Food and DrugAdministration), because it can be used to determine clinically rele-vant information. As a result, while some products have interpretedcustomers’ genetic data for both traits and ancestry purposes (e.g.23andMe), others avoid regulatory issues by only analyzing ancestryaspects (e.g., AncestryDNA).

3Background

3.1 Historical context and new technologies

Prior to the rapid expansions in genetic technologies, genetic testingwas confined to a specialized aspect of clinical practice. Testing for thediagnosis of a suspected genetic disorder (like Huntington’s disease orcystic fibrosis) occurred only for affected individuals and their imme-diate family, in rare cases. The data and information produced wasvery specific, and was interpreted by a lab specialized in testing forthe particular disease. The information was conveyed to the patientby trained experts called genetic counselors. These models for infor-mation management have been challenged by developments in genomictechnology, which have enormously expanded both the scope of geneticdata available and the number of individuals tested.

As outlined above, developments in genetic technologies have vastlylowered costs, creating a massive expansion in the scope of data pro-duced by genetic testing. Analysis of this data is an area of ongoingdevelopment, as researchers and clinicians adapt and combine disease-specific and gene-specific analytical approaches. The ability to performcomprehensive analysis is also limited by current knowledge, as ourunderstanding of the effects of gene variants continues to evolve.

13

14 Background

Parallel to this expansion in the amount of data produced is anexpansion in the scope of individuals tested. While genome andexome technologies have seen increasing clinical applications [Manolioet al., 2013], lowered costs for testing have also led to the growth ofdirect-to-consumer genetic testing (DTCGT) companies [Turrini andPrainsack, 2016]. These companies have brought genetic testing outsidethe clinic to millions of users (e.g. [23andMe, AncestryDNA]), sellingonline products that enable individuals to access their own genetic dataonline for the purposes of expanding their understanding of personalancestry, health, and biology.

3.2 Online interaction with personal genomics

The expansion of both the amount of data and the scope of individ-uals involved in genetic testing has led to radical changes both in thenature of the information returned and in the process of returningdata. Automated methods and online tools have become necessaryfeatures for sharing extensive information with many individuals. Inaddition, the scale of the underlying data defies comprehensive interpre-tation, and scientific understanding of the data is constantly evolving —many potentially meaningful genetic variants are not yet known to bemeaningful. What can be understood from genetic data, the potentialfor re-interpretation, and how to share that knowledge have provokedextensive debates [Hughes, 2013, Conley, 2014].

DTCGT companies enable individuals to acquire genetic informa-tion without the involvement of a healthcare provider by sending asaliva sample to a DTCGT company, at a relatively low cost. To date,DTCGT do not offer whole genome or exome sequencing, but rathermicroarray-based genotyping. Results are delivered through onlineinteractive reports. Several popular DTCGT services offer interactiveonline reports of ancestry information (e.g. 23andMe, AncestryDNA,and FamilyTreeDNA). The service 23andMe also provided risk assess-ment results for about 250 traits and conditions; however, in response toconcerns from the FDA current 23andMe health-related reports return

3.3. Omic technology 15

a more limited list of FDA-approved analyses [Pollack, 2013, 2015] (seeFigure 3.1). Services also allow users to share and compare genetic datawith each other, enabling comparison reports related to ancestry andgene inheritance (See Figures 3.2 and 3.3).

To date, all of the DTCGT services mentioned above also returnraw genotyping data to users, who in turn can actively engage withtheir personal genomic data, for example, by learning about specificgene variants or conditions of interest. Indeed, consumers of genomicdata have been observed transporting their data between services tocapitalize on different features that allow them to engage more deeplywith their data. For example, 23andMe users may export their datato FamilyTreeDNA for genealogy, or to various genetic data analysistools [Bettinger, 2013]. Because this data is digital, and because itsinterpretation gets updated frequently based on new research findings,we anticipate increased focus on the development of online interactivereport methods that perform automatic reanalysis.

3.3 Omic technology

The revolution in genetic data is paralleled by other omic technolo-gies that bring similarly broad, untargeted biological data to clinicaland direct-to-consumer contexts, including microbiome and metaboliteprofiling [Khamsi, 2014, Seetharaman, 2015, Dillet, 2015]. Microbiomeprofiling uses high-throughput DNA sequencing to profile microbiotathat live in the human gut, as well as other tissues, potentially impact-ing digestion, weight, and other aspects of health. Metabolite profilingassays numerous small molecules in the blood, and can reveal abnor-malities in metabolism and health. Like genome analyses, these andother omic technologies produce large amounts of data relevant tomany potential conditions. It seems likely that these technologies willconfront parallel challenges with management of comprehensive data,potentially limited and shifting interpretations, and the demand fortools to communicate information to non-expert individuals.

16 Background

Fig

ure

3.1:

Anexam

pleof

the“Inh

erite

dCon

ditio

ns”repo

rtprod

uced

by23an

dMe,

screen

ingforserio

usgene

ticcond

ition

s.In

thisrepo

rttheuser

ispresentedinform

ationforeverytested

gene

andlocatio

n.Becau

sethesetestsareusua

llyne

gativ

e(no

disease-causingvaria

ntfoun

d),itiscommon

forusersto

have

over

50results

that

areentir

ely“V

ariant

absent”.Uncertainty

ininterpretatio

nof

varia

nteff

ects

isrepresentedby

aconfi

dencescore(num

berof

starsin

the“C

onfid

ence”column).

3.4. Genomic data privacy and access 17

Figure 3.2: DTCGT providers typically share information about ancestry, andsometimes information is provided for specific chromosome regions. For example,this 23andMe data comparison illustrates shared DNA segments in a pair of siblinggenetic data sets.

3.4 Genomic data privacy and access

Commercial companies typically provide their customers with access toraw genetic data. This raw data contains an individual’s gene variantinformation, without interpretation, and can potentially be reanalyzedusing third-party tools and resources. In contrast, research studies only

18 Background

Figure 3.3: In some cases, DTCGT companies may be able to share information formore complex data comparisons. For users that have genetic data from their grand-parents, 23andMe’s “family tree” tool can be used to discover which grandparentscontributed genetic material for specific genes.

rarely give their research subjects access to their raw genetic data,possibly due to legal concerns regarding regulation of genetic testing[Evans, 2014, Karow, 2014].

Privacy of genetic information has been a major concern for manyindividuals interested in genetic data analysis. Because it containsextensive ancestry and trait information, genetic data is potentiallyidentifiable (especially when combined with additional data), and in2013 a re-identification study demonstrated that inference of surnamesfrom Y-chromosome data could be used to re-identify dozens of “anony-mous” public research genomes [Gymrek et al., 2013]. Many individ-uals are concerned about discrimination in access to insurance basedon genetic data despite some legal protections (e.g. the United StatusGenetic Information Nondiscrimination Act of 2008) [EEOC, 2008;Sanderson et al., 2016]. Discrimination is still possible in other con-texts such as employment and in some cases can be impossible to reg-ulate. Genome data privacy has thus been a concern in both research

3.5. Public genomes and follow-up research 19

and commercial contexts [Nature, 2013, Phillips, 2015], and controlled-access sharing for genome data research is common (e.g. as performedby the United States Database of Genotypes and Phenotypes, dbGaP),while commercial groups often promise privacy for individual-levelgenetic data.

3.5 Public genomes and follow-up research

The Harvard Personal Genome Project (PGP) is a research study, ini-tiated in 2005 by George Church of the Harvard Medical School, thatseeks to improve understanding of how genomic and environmentalfactors influence human traits through the creation of a public dataset[Church, 2005]. PGP volunteers agree to share their genomic sequences,as well as health data, with the scientific community and the public.Today, more than 5,000 volunteers are enrolled in the project througha process of open consent [Lunshof et al., 2008, Ball et al., 2012, 2014]and have agreed to share their genomic information publicly. Of these,over 300 individuals have had whole genome sequencing data producedby the project and publicly shared.

Public genome data produced by the Harvard PGP is accompa-nied by a preliminary research report produced by the GET-Evidencesystem [Ball et al., 2012]. This report is given to participants prior topublic data release, to better inform them regarding the data they arechoosing to publicly share. Upon consent, these reports are publiclyshared alongside public genome data, as is the underlying database forvariant interpretations.

The Harvard PGP is distinguished from other public genomeresources in its ongoing work with participants. The PGP has histori-cally invited other research groups to work with its volunteers, enablingfollow-up research with this population and providing a unique oppor-tunity for studying HCI in genomic data. This process is now stream-lined and explicitly supported within the Open Humans platform, asuccessor project to the PGP. In Open Humans, researchers can recruitand work with PGP participants, as well as other members with other

20 Background

public and private genetic data such as microbiome, activity logs, GPS,and other diverse data sources [Open Humans].

3.6 Characteristics of personal genomic data

The data produced by genome sequencing, exome sequencing, and com-prehensive genotyping spans all ∼20,000 genes of the human genome.Comparing any two genomes reveals millions of genetic differences, andthe database of all observed variations in all humans currently exceeds150 million genetic variants [dbSNP, 2016]. The information that can belearned from this comprehensive data is diverse, complex, and evolving.

One of the most popular uses of this information is for ancestryanalysis. By comparing individual data with population-level informa-tion, analysis of DNA segments can predict racial and ethnic origins.More recent ancestry — the discovery of cousins and near relatives —is also possible through large databases that detect shared segmentsof DNA between individuals. In addition to ancestry, everything aboutourselves that is heritable is — in theory — contained within the data.Although many gene variants have disease associations that imply clin-ical utility, individuals are also motivated by curiosity and entertain-ment to learn about variants with no medical purpose (e.g. physicaltraits) [Vayena et al., 2012]. Analyzing data to produce health and traitinsights is daunting: ClinVar, a public database aggregating reportedeffects, now has over 100,000 gene variants. Aggregating and filteringresults to facilitate understanding of any individual genome is challeng-ing, and it is still unclear how to combine information from multiplegenes affecting a single trait or disease.

3.7 Uncertainty in personal genomics

The relationships between genes and health effects are in many casesnot well understood, and knowledge about such relationships evolvesdynamically with the development of new technologies, processes, andresearch results. Individuals are therefore often required to continu-ously reconsider their genomic information against the most currentresearch evidence. As a result, interaction with personal genomic data

3.7. Uncertainty in personal genomics 21

is unique compared to other forms of personal data, where the dynamicelement is often the data itself, which is usually sampled at intervalsover time with the objective of creating an incremental feedback loopto influence an individual’s behavior [Li et al., 2011]. In such tradi-tional forms of personal informatics, uncertainty often results from thecontext and accuracy of data tracking. Genomic data, on the otherhand, are largely stable during a person’s lifetime −99.9999% accuracyis typical, and continues to improve [Peters et al., 2012]. The uncer-tainty in personal genomics stems from the interpretation of the data,and from the evidence of related implications for the user’s health andtraits (see, for example, 23andMe’s “confidence” score, Figure 3.1). Thisform of uncertainty arises because our understanding of genetic vari-ants changes over time: hypotheses are corrected or updated, and newmedical research exposes new relationships between people’s geneticmakeup and their effects.

4Related Work

4.1 Personal informatics

Quantified Self, also known as personal analytics [Regalado, 2013] andpersonal informatics [Li et al., 2010], refers to research and practiceinvolving communities, practices, and systems that help people collectand reflect on their personal information. The increasing availability oflow-cost sensors has accelerated the practice of self-tracking as well asthe rise of the Quantified Self movement [Quantified Self]. Researchersand designers have developed and studied numerous self-tracking tech-nologies and applications for health and wellness [Klasnja and Pratt,2012, Swan, 2009]. An underlying assumption driving the growth ofthis field is that individual’s knowledge of their data facilitates reflec-tion, which in turn leads to self-discoveries and to lifestyle changes.While researchers of personal informatics and Quantified Self acknowl-edge the value of self-tracking technologies [e.g., Bentley et al., 2013,Consolvo et al., 2008, Lin et al., 2006, Mamykina et al., 2008, Patelet al., 2012], they also discovered several barriers toward the adoptionand effective use of self-tracking technologies including data integration

22

4.2. Communicating uncertainty 23

and interpretation, unsuitable visualization and analytics tools, poorskills for analyzing data, and fragmented data scattered across multipleplatforms [Bloss et al., 2010, Choe et al., 2015]. It is important to notethat most of the research identifying practices and barriers in personalinformatics has been conducted with expert users, quantified-selfers,who are early adopters, health enthusiasts, or patients.

Personal Genomics shares the main goals and assumptions ofpersonal informatics — facilitating self-discovery based on personalinformation. In the popular media, personal genomics is viewed as anextension of quantified self: “It’s the ultimate quantified-self set of num-bers. Forget how many steps you average in a week, or what your rest-ing heart rate is — DNA genotyping allows us to create the completespec-sheet for our bodies” [Lamkin, 2016]. However, personal genomicsis different from other self-tracking data in being largely stable duringa person’s lifetime while its interpretation, and related implicationsfor the user’s health, are dynamic. In addition, while personal genomicinformation is inherently personal, it is also shared among family mem-bers, thus affecting the health and wellbeing of its owner as well as offamily members and future offspring. Furthermore, personal genomicinformation could lead to discoveries about one’s ancestry that couldimpact not only one’s identity and sense of belonging but also familiesand communities.

Considering the complex and sensitive nature of personal genomics,its scale, and the dynamic nature of its interpretation, we view personalgenomics as a new frontier for personal informatics.

4.2 Communicating uncertainty

One of the key tasks of a personal genomic report is to communicateto its reader the uncertainty that is associated with their data and itsimplications. An abundance of work has investigated the visualizationof uncertain information in other domains.

Existing taxonomies for communicating uncertainty identify sourcesof uncertainty (and visual presentation techniques) [Skeels et al., 2008,Taylor and Kuyatt, 1994, Thomson et al., 2005]. Additional work

24 Related Work

explores cognitive biases of decision-making under uncertainty and cor-rective visual approaches [Inbar, 2007, Tversky and Kahneman, 1974].

In particular, researchers explored public perception and communi-cation of weather forecast uncertainty, showing that the general publicunderstands that uncertainty is inherent in weather forecasts as wellas some of the factors that increase uncertainty [Morss et al., 2008,Joslyn and Savelli, 2010, Joslyn et al., 2013]. The research shows thatwhen communicated carefully, uncertainty in forecast leads to betteroutcomes such as increased precautionary action for extreme weatherevents [LeClerc and Joslyn, 2012, Savelli and Joslyn, 2012]. Researchalso indicates that communicating uncertainty in an interface or visu-alization of weather forecast could increase trust and help people makebetter decisions [Gigerenzer et al., 2005, Roulston et al., 2006, Joslynand LeClerc, 2013]. Evidence from other application domains such asremaining driving range in electric cars [Jung et al., 2015] and weightscale [Kay et al., 2013] indicates that not including uncertainty infor-mation can decrease trust in the application.

Numerous applications tracking new types of personal and oftenuncertain data have explored how to present the data to encouragebehavior change and reflection [Rooksby et al., 2014, Epstein et al.,2015, Choe et al., 2015, Lee et al., 2015]. A study comparing visu-alizations of uncertainty found that participants’ judgment of thesevisualizations was significantly influenced by familiarity, ease of under-standing, and visual appeal [Greis et al., 2016]. Another study com-paring the impact of various representations of uncertainty on differentactivities concluded that different types of visualizations lead to differ-ent learning outcomes and suggested that an interactive display may bebest for communicating uncertain information [Nadav-Greenberg et al.,2008]. Similarly, a different study shows that the amount of presenteduncertainty leads to different outcomes of risk taking within a gamecontext [Greis et al., 2016].

This growing body of work is important for researchers to drawupon when studying how to communicate uncertainty of personalgenomic information.

4.3. Interaction with biological data sets 25

4.3 Interaction with biological data sets

To date, little HCI research has focused on direct user engagement withpersonal genomic information. Instead, researchers focused on develop-ing new ways of interacting with large-scale and complex biological datasets for use as a platform to explore novel interaction techniques, suchas tangible interaction [Shaer et al., 2013, Manshaei et al., 2016]. Sys-tems developed include a tangible interface for scientists designing newDNA molecules [Schkolne et al., 2004], a tabletop tangible interface forsystem biologists [Wu et al., 2011], and several tabletop interfaces forinteractive visualization of biological data sets in informal and formallearning settings, such as DeepTree [Block et al., 2012] and PhyloGenie[Schneider et al., 2012]. G-nome Surfer [Shaer et al., 2011] is a tabletopinterface for collaborative exploration of genomes; Tangible mtDNA isan active tangible and tabletop system for collaborative exploration ofmitochondrial DNA sequencing data in breast cancer patients [Man-shaei et al., 2016]. However, these systems were not designed to sup-port non-expert users in the self-exploration of their own genomic databut rather to support researchers and learners in exploring genomicdata sets.

4.4 DIY genome analysis

Individuals can potentially explore and analyze their genetic data in avariety of ways including: exploring information provided by the pri-mary testing service; secondary analyses performed by third parties; orpersonal analyses drawing on various resources and databases.

However, information from primary testing services is often the onlyinformation explored. In some services this information can be quiteextensive. 23andMe, for example, provides in-depth reports on each ofthe variants included in the health and trait analyses. These reportssummarize current research, population statistics, and provide links toprimary literature. Due to regulatory concerns, however, many othertesting providers avoid providing information regarding health andtraits associated with genetic variants. Even services that do providethis information (e.g. 23andMe) are still limited to the set of variants

26 Related Work

Table 4.1: A selection from a growing variety of products giving consumers toaccess to genetic data, as well as analysis tools and resources for interpretation.

Tools, products,& resources Type(s) Notes23andMe — Genotyping testing service

— Ancestry interpretation

— Relative finding

— Health & trait interpretation

— Primary testing service

— Exportable raw data

— Over 1 million customers

AncestryDNA — Genotyping testing service

— Ancestry interpretation

— Relative finding

— Primary testing service

— Exportable raw data

— Over 2 million customers

ClinVar — Variant informationdatabase

— Aggregated reports

— Expert resource

— Public domain

Enlis Genomics Personal — Variant function analysis — Accepts data import

Exome AggregationConsortium (ExAC)

— Variant informationdatabase

— Variant frequencyinformation

— Share-alike license

FamilyTreeDNA — Genotyping testing service

— Ancestry interpretation

— Relative finding

— Primary testing service

— Accepts data import

— Exportable raw data

Gedmatch — Ancestry interpretation

— Relative finding

— Accepts data import

Genos — Exome sequencing — Primary testing service

— Exportable raw data

Illumina UnderstandYour Genome

— Whole genome sequencing

— Health & trait interpretation

— Primary testing service

— Exportable raw data

Interpretome — Ancestry interpretation

— Health & trait interpretation

— Data analysis software

(Continued)

4.4. DIY genome analysis 27

Table 4.1: (Continued)

Tools, products,& resources Type(s) NotesOnline MendelianInheritance in Man(OMIM)

— Gene information database

— Variant informationdatabase

— Curated summaries ofprimary literature

PubMed — Publications database — Search primary literature

Promethease — Health & trait interpretation — Accepts data import

SNPedia — Variant informationdatabase

— Crowdsourced

— Non-commercial license

Veritas Genetics — Whole genome sequencing

— Health & trait interpretation

— Primary testing service

— Exportable raw data(added fee)

they are reporting upon. Thus, many individuals are interested in addi-tional analyses of raw genetic data from primary providers.

There are a variety of third-party services and tools for analyzinggenetic data, which provide tools for ancestry analysis or health/traitanalysis (see Table 4.1). Services like FamilyTreeDNA and Gedmatchaccept third-party data for ancestry analysis, and enable relative find-ing within their communities. Health and ancestry analyses are alsopossible using third-party tools like Enlis, Interpretome, and Prometh-ease. These tools often provide links to resources and databases con-taining variant function predictions and reported effects (e.g. ClinVar,ExAC, OMIM, SNPedia, PubMed).

Table 4.1 presents a selection of primary testing providers, as wellas third-party analysis services, tools, resources, and databases usedfor exploring personal genomic information. For each of these we sum-marize its main features.

The rest of this article lays the ground for future HCI research ondirect engagement of non-experts with personal genomics.

5Emerging Framework for HCI for Personal

Genomics

In this section, we present a framework for HCI research for personalgenomics. The framework provides a research agenda, which inves-tigates the role of HCI methodology and interventions in personalgenomics. We center our framework on three themes — Understanding,Informing, and Empowering users:

Understanding users — examining users’ motivations, needs, informa-tion practices, and concerns in engaging with and sharing their personalgenomic information;

Informing users — investigating how variations in user interface designaffect users’ deliberation on their consent to genetic testing, their com-prehension of genomic information, as well as their intentions to sharesuch information with others;

Empowering users — focusing on the design and development of inter-active tools that empower users to engage with their personal genomicinformation over time in applicable and meaningful ways.

For each theme we review the recent and current research, and highlightopen questions for future HCI research.

28

5.1. Understanding users 29

5.1 Understanding users

Since the technology that enables lay people to interact with per-sonal genomic information directly is relatively new, many questionsremain open regarding non-expert engagement with personal genomicinformation.

For example, what motivates non-expert individuals to acquire per-sonal genomic information? What concerns do they have about engag-ing with personal genomics? What are their goals in exploring theirpersonal genomics? How do they learn from their personal genomicinformation? Which tools do they use and how do users document andorganize their findings? How do they validate their findings? What kindof information do they share and with whom?

Gaining insights into individuals’ motivations, needs, informationpractices, and concerns of engaging with and sharing their personalgenomic information plays a vital role in the design of new inter-active tools that empower non-experts to learn from their personalgenomic information. However, the importance of understanding per-sonal genomic users goes beyond the development of new tools toinforming future policy about personal genomics as well as to envi-sioning the future of personal informatics.

5.1.1 Current research on understanding users

Several studies have investigated the motivation of DTCGT users. Inthese studies, curiosity was mentioned as the participants’ primarymotivation for undergoing genomic testing [Goldsmith et al., 2012].Most participants wanted to learn more about themselves, were curiousabout their genetic makeup, or wanted to learn about individual geneticrisk factors. Participants also stated that they would use informationgained from the test to take personal responsibility for their futurehealth [McGowan et al., 2010]. Other themes included fascination withgenealogy, contribution to research, and recreation [Goldsmith et al.,2012]. Studies also identified several concerns among DTCGT users,including privacy, as well as the nature of the results and their futureimpact [Kuznetsov et al., 2015]. For example, users may develop anxiety

30 Emerging Framework for HCI for Personal Genomics

surrounding the risk of particular conditions, the possibility of discrim-ination from employers and insurance companies, or the possession ofsensitive and personal data. The Impact of Personal Genomics Study[PGen, 2016] is a current large-scale longitudinal study that surveysconsumers of two U.S. DTCGT providers — 23andMe and PathwayGenomics — to determine the characteristics of consumers, the psy-chological, behavioral, and health impact of genetic testing, and theethical, legal, and social issues associated with DTCGT services.

Little empirical data exists about the attitudes and motivationsof people who have had their whole genome sequenced [Goldsmithet al., 2012, Linderman et al., 2016], since a relatively a small num-ber of users around the world have had their entire genome sequencedand returned to them. Of the individuals that have access to theirwhole genome information, a significant subset have been volunteers ofthe Harvard Personal Genome Project (PGP). We established a designpartnership with the Harvard PGP and have collaborated closely withits researchers and volunteers to study the information practices andneeds of personal genomics users. PGP volunteers can be categorizedas early adopters [Rogers, 2003] or extreme user group [Choe et al.,2014, Troshynski et al., 2008]. They tend to have advanced educationand possess favorable attitudes toward science. Their motivations andinformation practices might not be generalizable or applicable to thebroader population. However, in areas that evolve rapidly based ontechnical and scientific innovations, the perspectives of early adoptersprovide important insights, because they have used the existing tech-nologies and have had the opportunity to identify challenges and poten-tial solutions [Choe et al., 2014, Troshynski et al., 2008].

To better understand the needs and information practices of PGPparticipants, we conducted a qualitative study with 63 participantsfrom the Harvard PGP volunteer community [Shaer and Nov, 2014].Participants filled out an online questionnaire that consists of 10open questions (see Shaer and Nov [2014]) regarding their motivation,goals, and information practices, in addition to demographics ques-tions. Although study participants were motivated by a diverse set ofgoals, ranging from understanding traits, to identifying health risks,

5.1. Understanding users 31

Figure 5.1: Five common information tasks of non-expert personal genomic users.

to learning about their ancestry, to contributing to science, they usedinteractive tools to perform mostly the five common information taskslisted in Figure 5.1. The study also identified high-level needs of non-experts seeking to learn more from their personal genomic data infor-mation. Figure 5.2 presents a summary of these needs.

From an HCI perspective, these findings can serve as a basistoward the design of new interaction techniques and tools for personalgenomics. To that end, it is important to gain further and more nuancedinsight into how users engage with, and learn from, their annotatedpersonal genomic reports. For that reason, we conducted a second qual-itative study of personal genomics users [Shaer et al., 2015]. We inter-viewed and observed 36 PGP participants, who have had their wholegenome sequenced, as they explored their personal genomic data usingthe GET-Evidence interactive genomic report [Block et al., 2012]. Thisstudy deepened our understanding of the needs and practices of per-sonal genomic users, highlighting that individuals are predominantly

32 Emerging Framework for HCI for Personal Genomics

Figure 5.2: Needs of non-experts seeking to learn from their personal genomic datainformation.

concerned with genetic variants that are well-established, pathogenic,and have high clinical importance.

While these insights are important for the development of inter-active tools and reports, important questions about user engagementwith personal genomics remain open.

5.1.2 Open questions for future research

In this section, we highlight two open questions for future HCI researchon understanding users:

1. Perceiving uncertainty and discrepancies: Although personalgenomic data are largely stable during a person’s lifetime, theirinterpretation and implications might change over time as newmedical research exposes relationships between genes and health.In order to accurately communicate the inherent uncertainty ofpersonal genomic interpretations, HCI research should investigate

5.1. Understanding users 33

how users of personal genomics perceive uncertainty and discrep-ancies between their personal genomic information, their expe-riences, and their family history. Such investigations could drawupon research the rich body of literature examining how lay indi-viduals understand uncertainty in other domains (e.g. weatherforecasting), which we discuss in Section 4.

2. Motivation for curating and sharing information: While per-sonal genomic information is personal, it is also shared amongfamilies and community members. Curating and sharing find-ings from a personal genomic report could motivate familiesto seek the advice of healthcare providers, promote scientificunderstanding of relationships between genes and traits, enablebiomedical research into rare conditions, and bring together com-munities of shared origins or interests. Additional research isrequired to understand how people curate personal genomic andrelated information, what factors influence people to share per-sonal genomic information, what information they share with oth-ers, and with whom they share such information. Such researchcould be informed by the increasing amount of literature on fam-ily informatics [Colineau and Paris, 2011, Grimes et al., 2009,Pina et al., 2017] and on engaging patients and their family care-givers in perusal of patient-provided data and clinical data, e.g.[Grimes et al., 2009, Colineau and Paris, 2011, Chung et al., 2016,Woollen et al., 2016, Pina et al., 2017]

Qualitative HCI research methods including longitudinal semi-structured studies [Blandford et al., 2016] have an essential role inaddressing these questions. However, when choosing a research method,careful attention should be directed toward considering the ethicalimplications of the study. Due to the sensitive nature of genetic test-ing, researchers should consider a range of ethics and privacy issuesranging from asking participants to share private information, to gath-ering sensitive information from social media and discussion groups,to presenting users with new interpretations of their personal genomicinformation that could drastically impact their understanding of theirhealth risks, identities, and of their families.

34 Emerging Framework for HCI for Personal Genomics

5.2 Informing users

The decreasing costs of obtaining personal biological data, coupled withthe proliferation of websites and digital devices for self-tracking andtesting, have resulted in a reduced barrier to entry for participation inonline biomedical research in general, and in genetic testing in partic-ular, and have highlighted the importance of informed online consentprocesses. Taken together, these factors require us to re-evaluate theeffectiveness and potential of digital, interactive consent in this newcontext of enrolling in biomedical research.

From an HCI perspective, interactive consent poses several chal-lenges and opportunities to informing users: whereas individuals par-ticipating in biomedical research in general or in genetic testing inparticular were formerly able to engage with a professional in a face-to-face dialogue, potential online research or DTCGT participants havefewer opportunities to ask questions and express their concerns in realtime [Balestra et al., 2015]. Furthermore, presentation techniques anddesign interventions may influence an individual’s decision to partici-pate [Friedman et al., 2000, Das et al., 2015], raising concerns aboutthe voluntariness of participation. In response to such concerns, federalagencies are drafting guidelines for electronic consent [FDA, 2015]. Atthe same time, interactivity offers unique advantages to the consent-seeking and deliberation process: through interactive consent forms,consent seekers may provide additional on-demand information basedon participants’ interests; through careful use of feedback and user ana-lytics, consent seekers may also be able to gain insights into the partsof the consent forms that participants find more useful, more chal-lenging, or requiring additional clarification. Yet another way in whichweb-based interactive consent forms can help the consent-seeking anddeliberation process includes the use of social features, such as rat-ings, recommendations, and annotations. Such features could enableindividuals deliberating on their consent decision to share information,evaluate different perspectives, and ultimately explore the risks andbenefits of the research beyond the scope of one-on-one dialogue witha research staff member.

5.2. Informing users 35

In addition to facilitating an informed consent process, HCI hasan important role in enhancing non-expert comprehension of personalgenomic reports, and in particular to help individuals make sense ofthe uncertainty of personal genomic data and its implications.

5.2.1 Current research on informing users

5.2.1.1 Improving online consent processes

The decision to consent to participate in medical research is medi-ated by two main factors: participants’ comprehension of the details ofthe study, and their trust in the research organization [Kneipp et al.,2009]. Recent studies on consent form design focus predominantly onthe impact of content structure, graphical enhancements, and multi-media on comprehension [Dresden and Levitt, 2001, Dunn et al., 2002,Murphy et al., 2007, Stiles et al., 2001].

To understand how social features can enhance online consentforms, we studied the use of social annotations in personal genomicresearch [Balestra et al., 2016a,b]. First, we explored the impact ofsocial annotation, embedded in online consent forms, on individuals’informed consent beliefs and decisions. Participants were presentedwith an online consent form for a personal genomics study, and wererandomly assigned to either a social annotation condition that exposedthem to previous users’ comments on-screen (Figure 5.3), or to a tra-ditional consent form without annotation. We compared participants’perceptions about their consent decision, their trust in the organiza-tion seeking the consent, and their consent decision across conditions.We found that while consent rates did not differ across conditions, onaverage individuals exposed to social annotation felt that their decisionwas more informed, and furthermore, that the effect of the exposure tosocial annotation was stronger among users characterized by relativelylower levels of prior privacy-preserving behaviors.

In a second study [Balestra et al., 2016b], we focused on the influ-ence of annotations’ valence on participants’ perceptions and behaviorssurrounding online consent for biomedical research: participants werepresented with an online consent form for a personal genomics study

36 Emerging Framework for HCI for Personal Genomics

Fig

ure

5.3:

Aconsentform

forpe

rson

algeno

micsresearch

with

social

anno

tatio

n,from

[Balestraet

al.,2016a,b].U

sers

can

read

comments

ontherig

ht,a

ndhigh

light

text

toleavetheirow

ncomments.

5.2. Informing users 37

that contained social annotations embedded in its margins. Individu-als were randomly assigned to view the consent form with positive-,negative-, or mixed-valence comments alongside the text of the con-sent form. We compared participants’ understanding of the materialand perceptions of informedness, their trust in the organization seek-ing the consent, and their consent across conditions. We found thatconsent forms containing positive-valence annotations are likely to leadparticipants to feel less informed and simultaneously more trusting ofthe organization seeking consent. In certain cases where participantsspent little time considering the content of the consent form, partici-pants exposed to positive-valence annotations were even more likely toconsent to the study.

5.2.1.2 Transforming personal genomic reports to a long-terminformation partnership

Traditionally, informed consent is a one-time event in which a personagrees to participating in research and is informed of what is knownto the science at that point in time. However, as medical research,and information about the relationships between genetic and healthinformation develop, digital informed consent offer opportunities forlong-term information partnerships between scientists and users. Insuch partnerships, the person consenting can be continuously informedof new knowledge about risks and benefits emerging from research,and the ensuing new interpretations of their genetic data. Further-more, by combining the nature of genetic information as data-staticbut interpretation-dynamic with design that incorporates social fea-tures and curation, genetic data can be curated by the users or rel-evant others (including healthcare professionals, family members, orothers with exposure to the user’s genetic data), such that emerg-ing new research information is shared with the user based on theirgenetic profile. Such approach turns a one-time consent event to anongoing informing mechanism which the user can enhance individuallyand socially. It also, however, represents potential risks of people con-senting to one thing (based on existing medical knowledge at the timeof consent) and being potentially exposed to information about other

38 Emerging Framework for HCI for Personal Genomics

things (based on advances in research). This may, in some cases, bemore stressful or mentally taxing than these people had anticipatedat the time of consent. Mechanisms can be designed for such cases,for example, by making the consent time-specific or with placing thecontrol of information sharing extent with the user.

5.2.1.3 Increasing comprehension of personal genomic reports

Existing research has focused on participants’ comprehension of anony-mous sample personal genetic reports. Lachance et al. [2010] examinedthe informational content, literacy demands, and usability of DTCGTservice websites. They found that websites vary widely, and most userswould struggle to use these resources effectively. The authors suggestedthat future tools focus on distilling and prioritizing important informa-tion while considering readability and usability elements. Other stud-ies have looked more specifically at users’ comprehension of genomicreports. Ostergren et al. [2015] assessed participants’ comprehensionof anonymized genomic reports and found that comprehension var-ied widely according to demographic characteristics, numeracy andgenetic knowledge, and types and format of the genetic informationpresented. They suggested that the presentation of genomic data be tai-lored to the test type and to consumer characteristics. To investigatethe effect of different visualizations on consumers’ understanding ofpersonal genomic data, we conducted a comparative study, which indi-cated an advantage to non-zoomable visualizations, with best results(in terms of both objective comprehension and subjective preference)from using bubble graphs [Shaer et al., 2015].

In contrast to the studies, which presented users with anonymizedgenetic data, Kuznetsov et al. [2015] presented users with their own23andMe data to understand how they make sense of and contextu-alize their results, critique and evaluate the underlying research, andconsider the broader implications of genetic testing. They framed con-sumers as members of biocitizen publics in which there is an empha-sis on individuals’ engagement with the community and higher-orderlearning processes [Airasian et al., 2001], rather than merely perceiving

5.2. Informing users 39

results and individually gathering information. The authors recom-mend the development of platforms for aggregating hybrid knowledgefor creative reflection on professional science, and for supporting col-laborations across communities.

5.2.1.4 Communicating uncertainty

The personal genomics context offers a form of uncertainty notaddressed by existing taxonomies and applications. In the genomicscontext, unlike most personal informatics contexts, the full data setis known and is mostly stable — the source of uncertainty is theinterpretation of the data, which depends on new scientific findings.Furthermore, personal genomics requires the communication of multi-dimensional uncertainty — uncertainty that emerges from the accumu-lation of several levels of uncertainty that the user encounters. Considerfor example cases where known relationships between genetic charac-teristics and medical conditions convey a probabilistic message (e.g.that the carrier of a certain genetic variant has a 10% likelihood ofdeveloping a certain medical condition). On top of that uncertainty,consider cases where the data supporting this relationship between thevariant and the medical condition is not well-established empirically, orthat it is based on studies that have not been replicated. Such accumu-lation of even two sources of uncertainty is highly complicated for mostnon-expert users, and more research is needed in order to understandhow to convey such information to non-experts in understandable ways.Karczewski et al. [2012] stress the importance of teaching the methodsof genetic risk calculations.

5.2.2 Open issues for future research

The following are some of key issues to be addressed by future HCIresearch on informing users:

1. Enhancing the online consent process with social features: whilerecent research indicates that adding social annotation to onlineconsent forms could potentially inform participants and improvetheir deliberation processes, many research questions remain

40 Emerging Framework for HCI for Personal Genomics

open. For example, what are the most effective ways to solicituseful content for social annotation? How, whether, and towhat extent should the content contributed by non-expert usersthrough social annotation tools be curated and edited before is itpresented to readers of informed consent forms? If such curationtakes place, who should be in charge of it? To what extent shouldthe social intervention be based on explicit user choices — forexample, should the exposure to social annotation be based onopt-in versus opt-out design defaults? What other social inter-ventions can be considered for an interactive informed consentform? And, finally, what are the ethical concerns associated withinterventions in consent forms, and what are the responsibilitiesof the consent seekers in this respect?

2. Personalized presentation and user choice: existing research sug-gests that the presentation of genomic information be tailored tothe test type and to the characteristics of the user. However, moreresearch is required to understand what elements should be inter-active and/or customized and in what ways. What are the bestpractices for design interventions that are tailored, transparent,and valuable to the user? How much autonomy should the userreceive to make sure they are well-informed, and to what extentshould design interventions be based on explicit user choices?

3. Communication of multidimensional uncertainty: delivering per-sonal genomic information to users requires the communication ofmultidimensional uncertainty — that is, uncertainty that emergesfrom the accumulation of several dimensions of uncertainty theuser encounters. Communicating such accumulating uncertaintyis not currently well addressed by existing taxonomies and tools,and poses a challenge to researchers and practitioners. Open ques-tions for researchers and practitioners include: How to best con-struct effective representations of multidimensional uncertainty?How should different levels of visual and statistical literacy beserved through interactive tools to facilitate multidimensionaluncertainty communication? And finally, how to evaluate users’

5.3. Empowering users 41

comprehension of multidimensional uncertainty? These investiga-tions could potentially draw upon the body of research on com-municating uncertainty described in Section 4, as well as on theproceedings of CHI 2017 workshop on Designing for Uncertaintyin HCI [Greis et al., 2017].

5.3 Empowering users

The applicability of personal genomic information to individuals, andits evolving understanding based on new scientific discoveries, evokecompelling HCI questions when considering digital tools that mediatethe interaction between the individual and their data: What are thefunctional requirements for supporting meaningful engagement withpersonal genomic information? How can we design tools which allowusers to explore and make sense of their personal genomic information?How can we design tools that support long-term engagement and per-sonal research? How can we empower users to contextualize and actupon their genomic information?

Interaction with personal genomic data often begins by reviewinga personalized genetic report [Shaer et al., 2015]. Thus, the designof interactive reports for non-expert individuals that support sense-making and exploration plays a vital role in empowering people tolearn from and engage with their personal genomic information. Self-exploration tools should also enable non-experts to contextualize andcompare their personal genomic information with other individuals(e.g. family members), ancestry information, and family medical his-tory [Kuznetsov et al., 2015] as well as to share information with a doc-tor, friends, family, or the public. Finally, since genetic research evolveswith the development of novel technologies, processes, and new scien-tific findings, tools for non-experts should support long-term engage-ment, allowing people to re-interpret their information in light of newscientific findings as well as personal developments.

5.3.1 Current research on empowering users

Our studies with personal genomic users indicate that non-expert indi-viduals mostly use the tools offered by their genetic data provider (e.g.

42 Emerging Framework for HCI for Personal Genomics

23andMe, PGP, AncestryDNA, 2016, fTree) [Shaer et al., 2014, Shaerand Nov, 2014]. However, tools offered by DTCGT companies pro-vide interpretations that reflect the preferences and limitations of thegenetic testing company and are mostly not customizable by the indi-vidual users. Typically, the results can only be generated and curatedby the DTCGT company. Also, since DTCGT tools may use propri-etary algorithms, their genome analysis is not always transparent to theuser and the user may not understand how the analysis was done or beable to replicate the results with other tools [Karczewski et al., 2012].Many of our study participants [Shaer et al., 2014, Shaer and Nov, 2014]have found the existing tools, both those offered by DTCGT providersand other online tools, to be too complicated and overwhelming interms of the amount of information they present and their use of sci-entific jargon.

To address these challenges, we presented a design case study ofa novel interactive tool, named GenomiX, which was developed usinga user-centered design process [Shaer et al., 2016]. It is important tonote that GenomiX does not provide new genome interpretations butrather draws upon the interpretation provided by the Harvard PGP’sGET-Evidence report. The requirements and design goals of GenomiXdraw upon findings from our previous research exploring users’ motives,needs, and interaction patterns with genomic data [Shaer et al., 2015].

Drawing upon these findings, we defined the following goals foran interactive tool for exploring personal genomic information: (G1)Presenting a visual summary of personal genomic information thathighlights which variants are potentially concerning and require fur-ther investigation; (G2) Communicating the level of certainty of thescientific evidence associating a particular gene variant to health con-ditions. Since the certainty of the evidence can change over time, thereport needs to provide up-to-date evidence. (G3) Relating variants tomedical conditions while conveying complex relations, which associatemultiple variants with a particular condition or the same variants withmultiple conditions. (G4) Allowing users to curate information aboutvariants, giving them a basis from which to conduct further research.

GenomiX is implemented as a web tool, presenting the user witha visualization that provides an overview of their genetic variant data

5.3. Empowering users 43

Figure 5.4: GenomiX: Gene Variant Report displaying the overview of a partici-pant’s results.

(Figure 5.4). Gene variants are represented as bubbles that are plot-ted between two axes: clinical importance (X axis), and certainty ofevidence (Y axis); the size, color, and placement of these bubbles com-municate specific information about each variant. Individuals can sortthe data on the plot by either risk or rarity of the variant. Alterna-tively, users can view a report organized by categories where variantsare sorted according to the anatomical system they might impact (seeFigure 5.5). Users can save information about particular gene variantsto a Notebook tab. GenomiX also offers contextual help regarding ter-minology as well as a glossary.

In addition to an interactive visual gene variant report, GenomiXoffers Curator, a tool designed to help users curate personal genomicinformation, make new connections, and engage with their personalgenomic information over time. Curator consists of two parts: a down-loadable Google Chrome extension, which allows users to save webpages, pdf files, videos, and images; and a Notebook, which displaysgene variants and web pages that a user has saved (Figure 5.6).

44 Emerging Framework for HCI for Personal Genomics

Figure 5.5: GenomiX: Gene Variant Report displaying participant’s results sortedby category.

Figure 5.6: The Notebook tool, displaying gene variants and web pages that a userhas saved.

5.3. Empowering users 45

To evaluate the usability and utility of GenomiX, we conductedthe first study of a new genome interpretation tool to date, in whichusers viewed their own personal genomic data [Shaer et al., 2016]. Ourfindings indicate that GenomiX offered its users a number of bene-fits afforded by the tool’s design features: visually reducing complex-ity improves users’ ability to prioritize gene variants, communicatinguncertainty helped users when interpreting genomic data and choosingwhich variants to focus on, and color coding helped users to identifyinformation they overlooked in prior tool such as distinguishing pro-tective or carrier variants.

While the design and evaluation of GenomiX offer insights into thedevelopment of future interactive personal genomics exploration tools,there are a number of key issues that should be considered in futureresearch about tools for empowering non-experts’ self-exploration ofpersonal genomics. Future research should also draw upon lessons froma rich body of research on using genetic data to empower patients inthe context of clinical care [e.g., Gilbar, 2007, Rosas-Blum et al., 2007,Berg et al., 2017]. Such lessons could be helpful in the design of noveltools for personal genomics exploration

5.3.2 Open issues for future research

Here, we highlight three key issues for future HCI research on empow-ering users:

1. Comparing and relating genomes: while existing tools presentusers with their own ancestry and allow relatives to view whichsegments of their DNA are shared (or not shared) between twoindividuals, additional research is needed to enable users tocompare and relate genomes in order to understand trait- andhealth-related information. Scientific methods to understand riskin the context of genome-sharing among family members arecomplex and are still evolving. Contextualizing genetic testingresults in respect to other family members or friends could playan important role in understanding and acting upon personalgenomic information. However, many questions remain open.What criteria for comparison are meaningful and informative for

46 Emerging Framework for HCI for Personal Genomics

non-experts? How to visualize multiple genome comparison fornon-experts in an effective manner? And how to facilitate shar-ing of information for comparison across individuals?

2. Facilitating sharing: from a societal perspective, informationsharing is an important aspect of users’ engagement with per-sonal genomic information. However, sharing such informationis rarely facilitated by existing tools, and is mostly limited toexploring ancestry and to discovering new biological relatives.Information sharing can be supported in a number of forms:

(1) Sharing personal genomic information: users can share theirgenomic information with other users (e.g. family mem-bers), with their healthcare provider, with research projectssuch as Harvard’s PGP, or with commercial efforts such as23andMe. In each one of these cases, users may choose toshare some parts of their information and withhold others;

(2) Sharing meta-information with social circles: this categoryincludes all information that is not the actual personalgenomic data, but rather related to it. The recipients ofthe information are the user’s social circles, online andoffline. Examples of such meta-information sharing include:the fact the user shared their information with a researchproject, the user’s thoughts about genomic information andthe role they play and can play in modern medicine, andthings the user learned as a result of engaging with theirpersonal genomic information. The value of this type ofinformation-sharing stems from its social impact: sharingsuch meta-information helps to increase awareness aboutthe opportunities and challenges associated with sharingpersonal genomic information;

(3) Sharing interpretations, insights, and other relevant infor-mation based on self and others’ personal genomic infor-mation. Examples of such information sharing include newscientific research findings related to personal genomic data

5.3. Empowering users 47

(of self or others) along with explanations of how the newresearch findings may be relevant to the genomic informa-tion, and what other relevant information might be neededas a next step in exploring it, or pointers for others whoshared their information about relevant online resources orcommunities (e.g. online communities of people with certainmedical conditions).

In all of the cases above, in addition to understanding users’ needsand concerns about sharing genomic information, it is importantto investigate which design and implementation features allowfor sharing while also supporting transparency regarding access,storage, and usage, and ensuring users control the level of privacythey prefer.

3. Evaluation methods — Personal genomic data are personal, andtherefore more studies are needed in which participants are pre-sented with their own data. Prior studies on user-facing genomicstools mostly use fictional or anonymized data; that is, genomictest results that do not belong to the participant. Thus, thesestudies fail to incorporate the impact and meaning of the datato the user whose data is reported. Also, existing studies tendto use novice participants who are unfamiliar with their genomicdata to evaluate visualizations, thereby being unable to look atparticipants’ evolving understanding of their genome, or howwell a proposed tool or platform provides new insights. Futurestudies should further investigate how new tools allow users toexplore and learn from their own data. In addition, tools for self-exploration of personal genomics have the potential to help usersmake sense of their genomic data over time as the research thatlinks this data to health outcomes evolves, and as users’ personalcircumstances change. Longitudinal studies are needed to exam-ine participants’ information needs and interactions over time andto understand how the use of personal genomic tools changes andhow to support users over time.

6Broader Implications

Making interaction with personal genomic information more accessiblefor non-experts, has broader implications for HCI researchers, practi-tioners, policymakers, and society.

6.1 For HCI researchers

Personal genomics presents a unique challenge to HCI scholars andpractitioners: while many contributions to HCI and personal informat-ics focus on novel ways to represent constantly-changing data, personalgenomics represents the opposite case: data is sampled only once and islargely static over the user’s lifetime, but the interpretation of the data,as well as its importance, can change dramatically over time, with newmedical research revealing hitherto unknown relations between geneticcharacteristics on one hand and physical attributes or medical condi-tions on the other. This combination of static information and dynamicinterpretation calls for new HCI approaches and tools that allow cap-turing dynamically and constantly changing new knowledge about astable data set. Such new tools may include dynamic curation of infor-mation related to genetic data, or adaptive decision support tools and

48

6.2. For practitioners 49

visualization, which change based on incoming new medical research. Arelated issue concerns the development of tools to facilitate sharing ofthis unique information–interpretation combination: future tools willneed to facilitate sharing with relevant others (e.g. family members,health providers, or scientists) not only personal genomic information,but also its changing interpretations, which may require the atten-tion — and in some cases action — of others. Finally, advances in HCIresearch may contribute to better communication of multidimensionaluncertainty by exploring the potential use of interactive, customized,and adaptive design.

6.2 For practitioners

HCI research can also contribute to the work of practitioners in areasrelated to personal genomics by exploring what information users valueand what information they are likely to be confused by. By studyinginformation behavior in a personal genomics context, we can improvethe ways in which genetic findings are communicated to users in clinicaland non-clinical settings. Products and tools based on HCI researchmay also underpin the development of an ecosystem of commercialand non-commercial services based on added layers of information, andon connecting people with relevant information, with relevant serviceproviders and with each other.

6.3 For society and policymakers

Better understanding and more effective engagement of users with theirpersonal genomic information can help increase genetic and health lit-eracy among non-experts. Moreover, rich and effective engagement mayalso lead to the further growth of burgeoning genomics-based citizenscience, in which members of the public contribute to professionalresearch projects as they collectively interpret and critique relevantinformation [Kuznetsov et al., 2015]. The accumulation of personalgenomic data and analysis contributed by members of the public cansupport the work of professional scientists in diverse areas such asgenetics, public health, and others.

50 Broader Implications

HCI work on personal genomics can also inform public discourseabout the relationship between people and their personal informa-tion, and in addition, between people and entities who can benefitfrom access to this personal information such as researchers, health-care providers, commercial companies, and public health organizations.As the scope of personal genomics grows and its effects on the publicbecome clearer, HCI work can inform regulators, policymakers, andstandard-setting institutions as they devise policies and regulations onhow personal genomics should be presented to the individuals (1) stati-cally, (2) interactively, and (3) dynamically. Recent events in which theFDA barred and then allowed personal genomics company 23andMeto offer personal genomic health-related reports to its users [Pollack,2015] exemplify some of the sensitivities and complexities involved inhow government can regulate the exposure of individuals to their per-sonal genomic data, and how interpretation of such data can or shouldbe provided to non-experts.

Finally, a societal aspect of the growing availability of DTCGTservices is that their customer base tends to gravitate toward whiteand well-educated users [Roberts and Ostergren, 2013]. It is there-fore our role as researchers to increase awareness and understanding ofgenetic information and its implications to populations that are under-represented in this user base. It should be also noted, that at a societallevel, enabling wide-spread understanding of the commonalities — andin particular, genealogical similarities — among people from differentethnic backgrounds may contribute toward reducing racial categoriza-tion and bias.

Acknowledgements

We would like to thank Christina Pollalis and Clarissa Verish for theircontribution. This work was partially funded by Grants IIS-1017693and IIS-1422706 from the National Science Foundation (NSF), Divisionof Information and Intelligent Systems (IIS).

51

References

23andMe. accessed 29 july 2016. https://www.23andme.com/, 2016.P. W. Airasian, K. A. Cruikshank, R. E. Mayer, P. R. Pintrich, J. Raths,

and M. C. Wittrock. A taxonomy for learning, teaching, and assessing: Arevision of bloom’s taxonomy of educational objectives. In L. W. Andersonand D. R. Krathwohl, editors, . Pearson, 2001.

AncestryDNA. accessed 29 July 2016. https://www.ancestry.com/dna/,2016.

M. Balestra, O. Shaer, J. Okerlund, M. Ball, and O. Nov. The effect of expo-sure to social annotation on online informed consent beliefs and behavior.Proceedings of the 19th ACM Conference on Computer-Supported Cooper-ative Work & Social Computing, pages 900–912, 2016a.

M. Balestra, O. Shaer, J. Okerlund, L. Westendorf, M. Ball, and O. Nov.Social annotation valence: The impact on online informed consent beliefsand behavior. Journal of Medical Internet Research, 18(7), 2016b.

M. P. Ball, J. V. Thakuria, A. W. Zaranek, T. Clegg, A. M. Rosenbaum,X. Wu, M. Angrist, J. Bhak, J. Bobe, M. J. Callow, C. Cano, M. F. Chou,W. K. Chung, S. M. Douglas, P. W. Estep, A. Gore, P. Hulick, A. Labarga,J. Lee, J. E. Lunshof, B. C. Kim, J. Kim, Z. Li, M. F. Murray, G. B. Nilsen,B. A. Peters, A. M. Raman, H. Y. Rienhoff, K. Robasky, M. T. Wheeler,W. Vandewege, D. B. Vorhaus, J. L. Yang, L. Yang, J. Aach, E. A. Ashley,R. Drmanac, S. Kim, J. B. Li, L. Peshkin, C. E. Seidman, J. Seo, K. Zhang,H. L. Rehm, and G. M. Church. A public resource facilitating clinical useof genomes. Proceedings of the National Academy of Science of the UnitedStates of America, 109(30):11920–11927, 2012.

52

References 53

M. P. Ball, J. R. Bobe, M. F. Chou, T. Clegg, P. W. Estep, J. E. Lunshof,W. Vandewege, A. Zaranek, and G. M. Church. Harvard personal genomeproject: Lessons from participatory public research. Genome Medicine, 6(2), 2014.

F. Bentley, K. Tollmar, P. Stephenson, L. Levy, B. Jones, S. Robertson,E. Price, R. Catrambone, and J. Wilson. Health mashups: Presenting sta-tistical patterns between wellbeing data and context in natural language topromote behavior change. ACM Transactions on Computer-Human Inter-action, 20(5), 2013.

J. S. Berg, P. B. Agrawal, D. B. Bailey Jr., A. H. Beggs, S. E. Brenner, A. M.Brower, J. A. Cakici, O. Ceyhan-Birsoy, K. Chan, F. Chen, R. J. Currier,D. Dukhovny, R. C. Green, J. Harris-Wai, I. A. Holm, B. Iglesias, G. Joseph,S. F. Kingsmore, B. A. Koenig, P. Y. Kwok, J. Lantos, S. J. Leeder, M. A.Lewis, A. L. McGuire, L. V. Milko, S. D. Mooney, R. B. Parad, S. Pereira,J. Petrikin, B. C. Powell, C. M. Powell, J. M. Puck, H. L. Rehm, N. Risch,M. Roche, J. T. Shieh, N. Veeraraghavan, M. S. Watson, L. Willig, T. W.Yu, T. Urv, and A. L. Wise. Newborn sequencing in genomic medicine andpublic health. Pediatrics, 2017.

B. Bettinger. What else can i do with my DNA test results? The GeneticGenealogist, accessed 29 July 2016, http://thegeneticgenealogist.com/2013/09/22/what-else-can-i-do-with-my-dna-test-results/, 2013.

A. Blandford, D. Furniss, and S. Makri. Qualitative HCI Research: GoingBehind the Scenes. Morgan & Claypool Publishers, 2016.

F. Block, M. S. Horn, B. C. Phillips, J. Diamond, E. Evans, and C. Shen. Thedeeptree exhibit: Visualizing the tree of life to facilitate informal learning.IEEE Transactions on Visualization and Computer Graphics, 18(2):2789–2798, 2012.

C. S. Bloss, L. Ornowski, E. Silver, M. Cargill, V. Vanier, N. J. Schork,and E. J. Topol. Consumer perceptions of direct-to-consumer personal-ized genomic risk assessments. Genetics in Medicine, 12(9):556–566, 2010.

E. K. Choe, N. B. Lee, B. Lee, W. Pratt, and J. A. Kientz. Understandingquantified-selfers’ practices in collecting and exploring personal data. InProceedings of the SIGCHI Conference on Human Factors in ComputingSystems, pages 1143–1152, 2014.

E. K. Choe, B. Lee, M. Kay, W. Pratt, and J. A. Kientz. Sleeptight: Low-burden, self-monitoring technology for capturing and reflecting on sleepbehaviors. In Proceedings of the 2015 ACM International Joint Conferenceon Pervasive and Ubiquitous Computing, pages 121–132, 2015.

54 References

C. F. Chung, K. Dew, A. Cole, J. Zia, J. Fogarty, J. A. Kientz, and S. A.Munson. Boundary negotiating artifacts in personal informatics: Patient-provider collaboration with patient-generated data. In Proceedings of theACM Conference on Computer-Supported Cooperative Work & Social Com-puting, pages 770–786, 2016.

G. M. Church. The personal genome project. Molecular Systems Biology, 1(1), 2005.

ClinVar. National Center for Biotechnology Information. accessed 29 July2016, http://www.ncbi.nlm.nih.gov/clinvar/, 2016.

N. Colineau and C. Paris. Family vs. individual profiles in a health portal:Strengths and weaknesses. In Proceedings of Conference on Human Com-puter Interaction, pages 321–330, 2011.

J. Conley. ACMG backs down a bit. Genomics Law Report, accessed29 July 2016, http://www.genomicslawreport.com/index.php/2014/04/07/acmg-backs-down-a-bit/, 2014.

S. Consolvo, D. W. McDonald, T. Toscos, M. Y. Chen, J. Froehlich, B. Har-rison, P. Klasnja, A. LaMarca, L. LeGrand, R. Libby, I. Smith, and J. A.Landay. Activity sensing in the wild: A field trial of ubifit garden. InProceedings of the SIGCHI Conference on Human Factors in ComputingSystems, pages 1797–1806, 2008.

S. Das, A. Kramer, L. Dabbish, and J. Hong. The role of social influence insecurity feature adoption. In Proceedings of the 18th ACM Conference onComputer Supported Cooperative Work & Social Computing, pages 1416–1426, 2015.

K. Davies. The $1,000 Genome: The Revolution In Dna Sequencing and theNew Era of Personalized Medicine. New York, NY: Free Press, 2010.

dbSNP. RELEASE: NCBI dbSNP built 148. National Center for Biotechnol-ogy Information, accessed 29 July 2016, http://www.ncbi.nlm.nih.gov/projects/SNP/snp_summary.cgi, 2016.

R. Dillet. Omixy improves medical check-up by making it aseasy as calling an uber. Tech Crunch, accessed 29 July 2016,https://techcrunch.com/2015/11/19/omixy-improves-medical-check-up-by-making-it-as-easy-as-ordering-an-uber/, 2015.

G. Dresden and A. Levitt. Modifying a standard industry clinical trial consentform improves patient information retention as part of the informed consentprocess. Academic Emergency Medicine, 8(3):246–252, 2001.

J. T. Dudley and K. J. Karczewski. Exploring Personal Genomics. Oxford,UK: Oxford University Press, 2013.

References 55

L. B. Dunn, L. A. Lindamer, B. W. Palmer, S. Golshan, L. J. Schneiderman,and D. V. Jeste. Improving understanding of research consent in middle-aged and elderly patients with psychotic disorders. American Journal ofGeriatric Psychiatry, 10(2):142–150, 2002.

EEOC. The genetic information nondiscrimination act of 2008. US EqualEmployment Opportunity Commission, 2014.

D. A. Epstein, A. Ping, J. Fogarty, and S. A. Munson. A lived informaticsmodel of personal informatics. In Proceedings of the 2015 ACM Inter-national Joint Conference on Pervasive and Ubiquitous Computing, pages731–742, 2015.

B. J. Evans. The first amendment right to speak about the human genome.Journal of Constitutional Law, 16(3):549–636, 2014.

FamilyTreeDNA. accessed 29 July 2016. https://www.familytreedna.com/,2016.

FDA. Use of electronic informed consent in clinical investigations questionsand answers, guidance for industry. US Food and Drug Adminis-tration, accessed 29 July 2016, http://www.fda.gov/downloads/drugs/guidancecomplianceregulatoryinformation/guidances/ucm436811.pdf&sa=D&ust=1469801013237000&usg=AFQjCNFJI8sZHaeA-Fiu6Rk7LENpHRrwKw, 2015.

B. Friedman, E. Felton, and L. I. Millett. Informed consent online: A con-ceptual model and design principles. University of Washington ComputerScience & Engineering Technical Report 00-12-2, 2000.

Genetic Information Nondiscrimination Act. https://www.eeoc.gov/laws/statutes/gina.cfm, 2008.

G. Gigerenzer, R. Hertwig, E. Van Den Broek, B. Fasolo, and K. V. Kat-sikopoulos. A 30% chance of rain tomorrow: How does the public under-stand probabilistic weather forecasts? Risk Analysis, 25(3):623–629, 2005.

R. Gilbar. Communicating genetic information in the family: The familialrelationship as the forgotten factor. Journal of Medical Ethics, 33(7):390–393, 2007.

L. Goldsmith, L. Jackson, A. O’Connor, and H. Skirton. Direct-to-consumergenomic testing: Systematic review of the literature on user perspectives.European Journal of Human Genetics, 20(8):811–816, 2012.

M. Greis, P. El.Agroudy, H. Schuff, T. Machulla, and A. Schmidt. Decision-making under uncertainty: How the amount of presented uncertainty influ-ences user behavior. Proceedings of the Nordic Conference on Human-Computer Interaction, 2016.

56 References

M. Greis, J. Hullman, M. Correll, M. Kay, and O. Shaer. Designing for uncer-tainty in HCI: When does uncertainty help? Extended Abstracts on HumanFactors in Computing Systems, 2017.

A. Grimes, D. Tan, and D. Morris. Toward technologies that support familyreflections on health. In Proceedings of the ACM International Conferenceon Supporting Group Work, pages 311–320, 2009.

M. Gymrek, A. L. McGuire, D. Golan, E. Halperin, and Y. Erlich. Identifyingpersonal genomes by surname inference. Science, 339(6117):321–324, 2013.

V. Hughes. It’s time to stop obsessing about the dangers ofgenetic information. Slate, Date accessed 29 July 2016, URL:http://www.slate.com/articles/health_and_science/medical_examiner/2013/01/ethics_of_genetic_information_whole_genome_sequencing_is_here_and_we_need.html, 2013.

O. Inbar. Graphical representation of statistical information in situations ofjudgment and decision-making. In Proceedings of 14th European Conferenceon Cognitive Ergonomics, pages 265–268, 2007.

S. Joslyn and J. LeClerc. Decisions with uncertainty: The glass half full.Current Directions in Psychological Science, 22(4):308–315, 2013.

S. Joslyn and S. Savelli. Communicating forecast uncertainty: Public per-ception of weather forecast uncertainty. Meteorological Applications, 17(2):180–195, 2010.

S. Joslyn, L. Nemec, and S. Savelli. The benefits and challenges of predictiveinterval forecasts and verification graphics for end-users. Weather, Climate& Society, 5(2):133–147, 2013.

M. F. Jung, D. Sirkin, T. M. Gür, and M. Steinert. Displayed uncertaintyimproves driving experience and behavior: The case of range anxiety in anelectric car. In Proceedings of the 33rd Annual ACM Conference on HumanFactors in Computing Systems, pages 2201–2210, 2015.

K. J. Karczewski, R. P. Tirrell, P. Cordero, N. P. Tatonetti, J. T. Dudley,K. Salari, M. Snyder, R. B. Altman, and S. K. Kim. Interpretome: Afreely available, modular, and secure personal genome interpretation engine.Pacific Symposium on Biocomputing, pages 339–350, 2012.

J. Karow. First newborn sequencing study gets FDA green light while othersstill await approval. GenomeWeb, accessed 1 January 2017, 2014.

M. Kay, D. Morris, M. C. Schraefel, and J. A. Kientz. There’s no such thingas gaining a pound: Reconsidering the bathroom scale user interface. InProceedings of the 2013 ACM International Joint Conference on Pervasiveand Ubiquitous Computing, pages 401–410, 2013.

References 57

M. Keshavan. Made in China: Direct-to-consumer whole genomesequencing for $875. MedCity News, accessed 1 January 2017,http://medcitynews.com/2016/01/made-in-china-direct-to-consumer-whole-genome-sequencing-for-895/, 2016.

R. Khamsi. Can gut DNA sequencing actually tell you any-thing about your health? Newsweek, accessed 29 July 2016,http://www.newsweek.com/2014/07/25/can-gut-dna-sequencing-actually-tell-you-anything-about-your-health-259348.html,2014.

P. Klasnja and W. Pratt. Healthcare in the pocket: Mapping the space ofmobile-phone health interventions. Journal of Biomedical Informatics, 45(1):184–198, 2012.

S. M. Kneipp, B. J. Lutz, and D. Means. Reasons for enrollment, the informedconsent process, and trust among low-income women participating in acommunity-based participatory research study. Public Health Nursing, 26(4):362–369, 2009.

S. Kuznetsov, A. Kittur, and E. Paulos. Biological citizen publics: Personalgenetics as a site of public engagement with science. In Proceedings of the2015 ACM SIGCHI Conference on Creativity and Cognition, pages 303–312, 2015.

C. R. Lachance, L. A. H. Erby, B. M. Ford, V. C. Allen, and K. A. Kaphingst.Informational content, literacy demands, and usability of websites offeringhealth-related genetic tests directly to consumers. Genetics in Medicine,12(5):304–312, 2010.

P. Lamkin. Personal genome testing: The ultimate quantified self. Ware-able, accessed 29 July 2016, http://www.wareable.com/health-and-wellbeing/personal-genome-testing-the-ultimate-quantified-self-2298, 2016.

J. LeClerc and S. Joslyn. Odds ratio forecasts increase precautionary actionin severe weather events. Weather, Climate & Society, 4:263–270, 2012.

M. K. Lee, J. Kim, J. Forlizzi, and S. Kiesler. Personalization revisited: Areflective approach helps people better personalize health services and moti-vates them to increase physical activity. In Proceedings of the 2015 ACMInternational Joint Conference on Pervasive and Ubiquitous Computing,pages 743–754, 2015.

I. Li, A. Dey, and J. Forlizzi. A stage-based model of personal informaticssystems. In Proceedings of the SIGCHI Conference on Human Factors inComputing Systems, pages 557–566, 2010.

58 References

I. Li, A. Dey, J. Forlizzi, K. Höök, and Y. Medynskiy. Personal informaticsand HCI: design, theory, and social implications. In Extended Abstracts onHuman Factors in Computing Systems, pages 2417–2420, 2011.

J. J. Lin, L. Mamykina, S. Lindtner, G. Delajoux, and H. B. Strub.Fish’n’steps: Encouraging physical activity with an interactive computergame. In P. Dourish and A Friday, editors, The 8th International Confer-ence on Ubiquitous Computing, pages 261–278. Springer, 2006.

M. D. Linderman, D. E. Nielsen, and R. C. Green. Personal genome sequenc-ing in ostensibly healthy individuals and the peopleSeq consortium. Journalof Personalized Medicine, 6(2):14, 2016.

J. E. Lunshof, R. Chadwick, D. B. Vorhaus, and G. M. Church. From geneticprivacy to open consent. Nature Reviews Genetics, 9(5):406–411, 2008.

L. Mamykina, E. Mynatt, P. Davidson, and D. Greenblatt. MAHI: Investiga-tion of social scaffolding for reflective thinking in diabetes management. InProceedings of the SIGCHI Conference on Human Factors in ComputingSystems, pages 477–486, 2008.

T. A. Manolio, R. L. Chisholm, B. Ozenberger, D. M. Roden, M. S. Williams,R. Wilson, D. Bick, E. P. Bottinger, M. H. Brilliant, C. Eng, K. A. Frazer,B. Korf, D. H. Ledbetter, J. R. Lupski, C. Marsh, D. Mrazek, M. F. Murray,P. H. O’Donnell, D. J. Rader, M. V. Relling, A. R. Shuldiner, D. Valle,R. Weinshilboum, E. D. Green, and G. S. Ginsburg. Implementing genomicmedicine in the clinic: The future is here. Genetics in Medicine, 15(4):258–267, 2013.

R. Manshaei, N. Baig, S. DeLong, S. Khayyer, B. East, and A. Mazalek.Exploring genetic mutations on mitochondrial DNA cancer data with inter-active tabletop and active tangibles. In Proceedings of the 2016 ACM onInteractive Surfaces and Spaces, pages 461–464, 2016.

M. L. McGowan, J. R. Fishman, and M. A. Lambrix. Personal genomics andindividual identities: Motivations and moral imperatives of early users. NewGenetics and Society, 29(3):261–290, 2010.

M. Molteni. Genos will sequence your genes — and help you sell them toscience. Wired, accessed 1 January 2017, https://www.wired.com/2016/12/genos-will-sequence-genes-help-sell-science/, 2016.

R. E. Morss, J. L. Demuth, and J. K. Lazo. Communicating uncertainty inweather forecasts: A survey of the U.S. public. Weather and Forecasting,23(5):974–991, 2008.

References 59

D. A. Murphy, D. Hoffman, G. R. Seage, M. Belzer, J. Xu, S. J. Durako, andM. Geiger. Improving comprehension for HIV vaccine trial informationamong adolescents at risk of HIV. AIDS Care, 19(1):42–51, 2007.

L. Nadav-Greenberg, S. L. Joslyn, and M. U. Taing. The effect of uncer-tainty visualizations on decision making in weather forecasting. Journal ofCognitive Engineering and Decision Making, 2(1):24–47, 2008.

Nature. Genetic privacy. Nature, 493:451, 2013.NHGRI. The cost of sequencing a genome. National Human Genome Research

Institute, accessed 29 July 2016, https://www.genome.gov/27565109/the-cost-of-sequencing-a-human-genome/, 2016.

NIH. Precision medicine initiative cohort program. National Instituteof Health, accessed 29 July 2016, https://www.nih.gov/precision-medicine-initiative-cohort-program, 2016.

Open Humans. Research and data. OpenHumans, accessed 29 July 2016,https://www.openhumans.org/research/, 2016.

J. E. Ostergren, M. C. Gornick, D. A. Carere, S. S. Kalie, W. R. Uhlmann,M. T. Ruffin, J. L. Mountain, R. C. Green, J. S. Roberts, and PGen StudyGroup. How well do customers of direct-to-consumer personal genomictesting services comprehend genetic test results? Public Health Genomics,18(4):216–224, 2015.

R. A. Patel, P. Klasnja, A. Hartzler, K. T. Unruh, and W. Pratt. Probingthe benefits of real-time tracking during cancer care. In AMIA AnnualSymposium Proceedings, pages 1340–1349, 2012.

B. A. Peters, B. G. Kermani, A. B. Sparks, O. Alferov, P. Hong, A. Alexeev,Y. Jiang, F. Dahl, Y. T. Tang, J. Haas, K. Robasky, A. W. Zaranek, J. H.Lee, M. P. Ball, J. E. Peterson, H. Perazich, G. Yeung, J. Liu, L. Chen,M. I. Kennemer, K. Pothuraju, K. Konvicka, M. Tsoupko-Sitnikov, K. P.Pant, J. C. Ebert, G. B. Nilsen, J. Baccash, A. L. Halpern, G. M. Church,and R. Drmanac. Accurate whole-genome sequencing and haplotyping from10 to 20 human cells. Nature, 487(7406):190–195, 2012.

PGen. Accessed 31 January 2016. http://www.genomes2people.org/pgen/,2016.

A. M. Phillips. Genomic privacy and direct-to-consumer genetics: Big con-sumer genetic data — what’s in that contract? 2015 IEEE Security andPrivacy Workshops, pages 60–64, 2015.

60 References

L. R. Pina, S. W. Sien, T. Ward, J. C. Yip, S. A. Munson, J. Fogarty, and J. A.Kientz. From personal informatics to family informatics: Understandingfamily practices around health monitoring. In Proceedings of the ACMConference on Computer-Supported Cooperative Work & Social Computing,2017.

S. Pinker. My genome, my self. New York Times, accessed 29 July 2016, http://www.nytimes.com/2009/01/11/magazine/11Genome-t.html, 2009.

A. Pollack. F.D.A. orders genetic testing firm to stop sellingDNA analysis service. New York Times, accessed 29 July 2016,http://www.nytimes.com/2013/11/26/business/fda-demands-a-halt-to-a-dna-test-kits-marketing.html?_r=1, 2013.

A. Pollack. 23andme will resume giving users health data. New York Times,accessed 29 July 2016, http://www.nytimes.com/2015/10/21/business/23andme-will-resume-giving-users-health-data.html, 2015.

Quantified Self. Accessed 29 July 2016. http://quantifiedself.com, 2016.A. Regalado. Q&A: Stephen Wolfram on personal analytics. MIT Technology

Review, 116(4), 2013.A. Regalado. For $999, Veritas Genetics will put your genome on a

smartphone app. MIT Technology Review, accessed 1 January 2017,https://www.technologyreview.com/s/600950/for-999-veritas-genetics-will-put-your-genome-on-a-smartphone-app/, 2016.

J. S. Roberts and J. Ostergren. Direct-to-consumer genetic testing and per-sonal genomics services: A review of recent empirical studies. CurrentGenetic Medicine Reports, 1(3):182–200, 2013.

E. M. Rogers. Diffusion of Innovations. 5th edition, Free Press, 2003.J. Rooksby, M. Rost, A. Morrison, and M. Chalmers. Personal tracking as

lived informatics. In Proceedings of the 32nd Annual ACM Conference onHuman Factors in Computing Systems, pages 1163–1172, 2014.

E. Rosas-Blum, P. Shirsat, and M. Leiner. Communicating genetic informa-tion: A difficult challenge for future pediatricians. BMC Medical Education,7(17), 2007.

M. S. Roulston, G. E. Bolton, A. N. Kleit, and A. L. Sears-Collins. A labora-tory study of the benefits of including uncertainty information in weatherforecasts. Weather and Forecasting, 21(1):116–122, 2006.

References 61

S. C. Sanderson, M. D. Linderman, S. A. Suckiel, G. A. Diaz, R. E. Zinberg,K. Ferryman, M. Wasserstein, A. Kasarskis, and E. E. Schadt. Motivations,concerns and preferences of personal genome sequencing research partici-pants: Baseline findings from the healthseq project. European Journal ofHuman Genetics, 24(1):14–20, 2016.

S. Savelli and S. Joslyn. Boater safety: Communicating weather forecast infor-mation to high stakes end users. Weather, Climate & Society, 4:7–19, 2012.

S. Schkolne, H. Ishii, and P. Schroder. Immersive design of DNA moleculeswith a tangible interface. In Proceedings of the Conference on Visualization,pages 227–234, 2004.

B. Schneider, M. Strait, L. Muller, S. Elfenbein, O.Shaer, and C. Shen. Phylo-genie: Engaging students in collaborative ‘tree-thinking’ through tabletoptechniques. In Proceedings of the SIGCHI Conference on Human Factorsin Computing Systems, pages 3071–3080, 2012.

D. Seetharaman. Jessica Richman seeks samples to crack thehuman microbiome. Wall Street Journal, accessed 29 July 2016,http://www.wsj.com/articles/jessica-richman-seeks-samples-to-crack-the-human-microbiome-1445452170, 2015.

O. Shaer and O. Nov. HCI for personal genomics. Interactions, 21(5):32–37,2014.

O. Shaer, M. Strait, C. Valdes, T. Feng, M. Lintz, and H. Wang. Enhanc-ing genomic learning through tabletop interaction. In Proceedings of theSIGCHI Conference on Human Factors in Computing Systems, pages 2817–2826, 2011.

O. Shaer, A. Mazalek, B. Ullmer, and M. Konkel. From big data to insights:opportunities and challenges for TEI in genomics. In Proceedings of the 7thInternational Conference on Tangible, Embedded and Embodied Interaction,pages 109–116, 2013.

O. Shaer, O. Nov, A. West, and D. Eastman. Understanding informationpractices of interactive personal genomics users. In Extended Abstracts onHuman Factors in Computing Systems, pages 2455–2460, 2014.

O. Shaer, O. Nov, J. Okerlund, M. Balestra, E. Stowell, L. Ascher, J. Bi,C. Schlenker, and M. Ball. Informing the design of direct-to-consumer inter-active personal genomics reports. Journal of Medical Internet Research, 17(6), 2015.

62 References

O. Shaer, O. Nov, J. Okerlund, M. Balestra, E. Stowell, L. Westendorf, C. Pol-lalis, J. Davis, L. Westort, and M. Ball. Genomix: A novel interaction toolfor self-exploration of personal genomic data. In Proceedings of the 2016Conference on Human Factors in Computing Systems, pages 661–672, 2016.

M. Skeels, B. Lee, G. Smith, and G. Robertson. Revealing uncertainty forinformation visualization. In Proceedings of Advanced Visual Interfaces,pages 376–379, 2008.

P. G. Stiles, N. G. Poythress, A. Hall, D. Falkenbach, and R. Williams.Improving understanding of research consent disclosures among personswith mental illness. Psychiatric Services, 52(6):780–785, 2001.

M. Swan. Emerging patient-driven health care models: An examination ofhealth social networks, consumer personalized medicine and quantified self-tracking. International Journal of Environmental Research and PublicHealth, 6(2):492–525, 2009.

B. N. Taylor and C. E. Kuyatt. Guidelines for evaluating and expressing theuncertainty of NIST measurement results. NIST Technical Note 1297, 1994.

J. Thomson, E. Hetzler, A. MacEachren, M. Gahegan, and M. Pavel. Typologyfor visualizing uncertainty. In Proceedings of the IS&T/SPIE Symposium onElectronic Imaging, Conference on Visualization and Data Analysis, pages146–157, 2005.

E. Troshynski, C. Lee, and P. Dourish. Accountabilities of presence: Refram-ing location-based systems. In Proceedings of the SIGCHI Conference onHuman Factors in Computing Systems, pages 487–496, 2008.

M. Turrini and B. Prainsack. Beyond clinical utility: The multiple values ofDTC genetics. Applied & Translational Genomics, 8(1):4–8, 2016.

A. Tversky and D. Kahneman. Judgment under uncertainty: Heuristics andbiases. Science, 185(4157):1124–1131, 1974.

E. Vayena, E. Gourna, J. Streuli, E. Hafen, and B. Prainsack. Experiences ofearly users of direct-to-consumer genomics in switzerland: An exploratorystudy. Public Health Genomics, 15(6):352–362, 2012.

J. Woollen, J. Prey, L. Wilcox, A. Sackeim, S. Restaino, S. T. Raza, S. Bakken,S. Feiner, G. Hripcsak, and D. Vawdrey. Patient experiences using aninpatient personal health record. Applied Clinical Informatics, 7(2):446–460, 2016.

A. Wu, J. B. Yim, E. Caspary, A. Mazalek, S. Chandrasekharan, and N. J.Nersessian. Kinesthetic pathways: A tabletop visualization to support dis-covery in systems biology. In Proceedings of the 8th ACM Conference onCreativity and Cognition, pages 21–30, 2011.

View publication statsView publication stats